Is there any way i can lower the amount of power dissipation through my power mosfets in my H-bridge

Thread Starter

rfengineer28

Joined Apr 28, 2021
71
1654144527529.png

Is it normal for me to expect this amount of power dissipation through my mosfets while moving 120V to control a motor? i have tried to address this issue but i keep seeing this amount of dissipation over and over again. If i higher the resistance at my gates i get a much lower gate current which is not what i want since i want to use the driver to its' full potential.
 

LowQCab

Joined Nov 6, 2012
2,157
The IR2110 can only supply around ~2-Amps-Peak, this is no where near enough Current.

You need something like 4 of these ...........
IXDN630MCI,
with some big high-quality Ceramic Bypass Capacitors attached close to each Driver, ( at least 2uf ).

They are available in Inverting, and Non-Inverting part-numbers for your convenience.
.
.
.
 

Attachments

Ian0

Joined Aug 7, 2020
5,473
This is a SPICE simulation - how well does it work in real life?
I think that the voltage and current waveforms on a much smaller timescale would be much more useful to be able to tell what is going on than SPICE's power dissipation estimates.
 

Papabravo

Joined Feb 24, 2006
18,770
I'm missing any source of pulse or switching waveforms on your schematic which is too small to see any details on. Separate schematic and waveform screenshots would be helpful. Putting that aside for a moment.

Switching losses come in at least two flavors:
  1. Rds(on) when the device is fully turned on. This number is given in the datasheet along with some parameters that tell how the measurement was made. You want this to be as small as possible. For a given device you pretty much have to take what you get.
  2. How much time do you spend on the "Miller Plateau"? This is the linear region where the MOSFET is not completely on or completely off. This is where your "hidden" losses come into play. You may be able to control these by choosing an appropriate gate driver to minimize the amount of time you spend here.
As @Alec_t mentioned, a provision for deadtime in your switching waveforms is essential. Nothing will kill your overall efficiency like briefly shorting the power supply 2000 times per second.

You can put an adjustable deadtime in your simulations and look for an actual part to implement it when you get them working. I used the "Complementary Buffer with Deadtime" from the LTspice Control Library, which you can find on GitHub or on the @Bordodynov website.
http://bordodynov.ltwiki.org/
It is the first hyperlink titled "Additional Library for LTspice..."

1654169681570.png
1654170050281.png

I think a BAHS heatsink would be appropriate for these guys, and yes I know that I've exceeded the maximum Vds for the IRHF5110. Simulations are made for the express purpose of keeping the magic smoke inside.

Lastly: Would you consider a change to your LTspice waveform color scheme? Dark Blue on Black just ain't makin' it. A white background works with most colors ( @Bordodynov & @ericgibbs ) and a black background works with bright red, yellow, and green ( @crutschow ).
 
Last edited:

Thread Starter

rfengineer28

Joined Apr 28, 2021
71
I'm missing any source of pulse or switching waveforms on your schematic which is too small to see any details on. Separate schematic and waveform screenshots would be helpful. Putting that aside for a moment.

Switching losses come in at least two flavors:
  1. Rds(on) when the device is fully turned on. This number is given in the datasheet along with some parameters that tell how the measurement was made. You want this to be as small as possible. For a given device you pretty much have to take what you get.
  2. How much time do you spend on the "Miller Plateau"? This is the linear region where the MOSFET is not completely on or completely off. This is where your "hidden" losses come into play. You may be able to control these by choosing an appropriate gate driver to minimize the amount of time you spend here.
As @Alec_t mentioned, a provision for deadtime in your switching waveforms is essential. Nothing will kill your overall efficiency like briefly shorting the power supply 2000 times per second.

You can put an adjustable deadtime in your simulations and look for an actual part to implement it when you get them working. I used the "Complementary Buffer with Deadtime" from the LTspice Control Library, which you can find on GitHub or on the @Bordodynov website.
http://bordodynov.ltwiki.org/
It is the first hyperlink titled "Additional Library for LTspice..."

View attachment 268582
View attachment 268585

I think a BAHS heatsink would be appropriate for these guys, and yes I know that I've exceeded the maximum Vds for the IRHF5110. Simulations are made for the express purpose of keeping the magic smoke inside.

Lastly: Would you consider a change to your LTspice waveform color scheme? Dark Blue on Black just ain't makin' it. A white background works with most colors ( @Bordodynov & @ericgibbs ) and a black background works with bright red, yellow, and green ( @crutschow ).
Hi , thanks for your response. The waveform for my simulation and the power dissipation accross the high side mosfet is displayed below.
1654201823128.png

would it help if i have a driver with a higher output current?
 

Thread Starter

rfengineer28

Joined Apr 28, 2021
71
What are the Hin and Lin parameters?
How much dead time are you allowing (the IR2110 doesn't inherently provide any/enough)?
I am running this simulation with a frequency of 4kHz and 50% duty cycle. I am setting Von as 5V and running it through a an amplifier of gain 4 as i need to increase the voltage from my arduino signal to the driver. Every time i intend to go in in 1 direction i set 1 amplifier to 5 and the other to 0 vice versa.

1654202025488.png
 

3Dogs

Joined May 5, 2020
1
99% of your simulated power consumption is occurring at the rising edge of your control signal.
This implies that you have shoot-through due to insufficient dead time.
Note that turning the FET on slower (using a gate resistor) makes everything worse. The only reason to use a gate resistor is for EMI/RFI reduction (slow fast edges down).
I’m not a SPICE user, but if you can zoom into the rising edge, and correlate that with your switching signal, it might help.
 

crutschow

Joined Mar 14, 2008
30,080
I agree that it looks like not enough dead time between the turning off of the top MOSFET before the bottom MOSFET is turned on.
Look at the expanded time between the two gate signals to see how much the dead-time is. (just hold the left mouse button to draw a box around the signal edges in the graph to zoom).
 

Thread Starter

rfengineer28

Joined Apr 28, 2021
71
I agree that it looks like not enough dead time between the turning off of the top MOSFET before the bottom MOSFET is turned on.
Look at the expanded time between the two gate signals to see how much the dead-time is. (just hold the left mouse button to draw a box around the signal edges in the graph to zoom).
1654292256603.png

This is pretty much whats happening, top side is my high side and bottom graph is my low side. will having proper dead time solve that issue?
 

crutschow

Joined Mar 14, 2008
30,080
For the high-side gate you need to measure the gate-source voltage.
You show that by left-clicking on the source node, and while holding down the button, move to the gate-node and release the button.

Please label the nodes you are measuring and show the schematic, so we know exactly what you are measuring.

And putting 20V on the low-side gate, as you show, may be marginal as most MOSFET gates have a maximum rating of 20V (or less).
 

Thread Starter

rfengineer28

Joined Apr 28, 2021
71
I'm missing any source of pulse or switching waveforms on your schematic which is too small to see any details on. Separate schematic and waveform screenshots would be helpful. Putting that aside for a moment.

Switching losses come in at least two flavors:
  1. Rds(on) when the device is fully turned on. This number is given in the datasheet along with some parameters that tell how the measurement was made. You want this to be as small as possible. For a given device you pretty much have to take what you get.
  2. How much time do you spend on the "Miller Plateau"? This is the linear region where the MOSFET is not completely on or completely off. This is where your "hidden" losses come into play. You may be able to control these by choosing an appropriate gate driver to minimize the amount of time you spend here.
As @Alec_t mentioned, a provision for deadtime in your switching waveforms is essential. Nothing will kill your overall efficiency like briefly shorting the power supply 2000 times per second.

You can put an adjustable deadtime in your simulations and look for an actual part to implement it when you get them working. I used the "Complementary Buffer with Deadtime" from the LTspice Control Library, which you can find on GitHub or on the @Bordodynov website.
http://bordodynov.ltwiki.org/
It is the first hyperlink titled "Additional Library for LTspice..."

View attachment 268582
View attachment 268585

I think a BAHS heatsink would be appropriate for these guys, and yes I know that I've exceeded the maximum Vds for the IRHF5110. Simulations are made for the express purpose of keeping the magic smoke inside.

Lastly: Would you consider a change to your LTspice waveform color scheme? Dark Blue on Black just ain't makin' it. A white background works with most colors ( @Bordodynov & @ericgibbs ) and a black background works with bright red, yellow, and green ( @crutschow ).
I havent been able to figure out how to put dead time into my circuit, i am currently switching at 4kHz.
 

Attachments

LowQCab

Joined Nov 6, 2012
2,157
Regardless of this particular aspect,
( which may be a very legitimate problem ),
the Thread-Starter is using some MASSIVE FETs and
trying to drive them with a max-Peak-Current of ~2-Amps.
.
Huge FET Ratings .PNG

In the real word that's not going to end well.
.
.
.
 
Top